Clinical nursing and postoperative prediction of gastrointestinal cancer based on CT deep learning model

To investigate nursing techniques and the implementation of CT deep learning based model in predicting postoperative prognosis of gastric cancer patients. Preoperative CT enhanced images and clinical data of 200 patients (134 in the training group and 66 in the validation group) with postoperative pathological diagnosis of gastric cancer were retrospectively collected. In addition, 51 cases of residual gastric cancer after subtotal gastrectomy were found, including 35 cases of radical total gastrectomy and 16 cases of palliative gastrectomy. Machine learning features were extracted from the maximum tumor layer of portal vein CT images. LASSOCox regression method was used to select features and construct labels. Then, Cox regression model was used to integrate labels and clinicopathological information to construct prediction models. Tumor segmentation was also performed to provide a more detailed analysis. There was a significant difference between the two surgeries, P < 0.05. All patients died within two years after palliative surgery, while no patients died after radical treatment. 10 deep learning features were selected from the images to construct deep learning labels. The labels were significantly correlated with overall survival time in both the training group and the verification group (P < 0.001 and P = staging showed good differentiation and calibration in training group [C-index (95%CI) = 0.776 (0.718–0.833)] and verification [C-index (95%CI) = 0.797 (0.680–0.914) ] CCC. Decision curve analysis shows that the prediction model has excellent diagnostics practicability. Deep learning model based on preoperative CT images can individually predict postoperative prognosis of gastric cancer patients, which is expected to assist clinical treatment decision-making and to assist in nuring of gastric patients. For patients after subtotal gastrectomy, prevention should be given priority to.